An independent analyst’s warning that Tesla’s Full Self-Driving safety metrics are “sharply deteriorating” has landed at a telling moment. The claim arrives just as federal regulators tighten the rules governing how automakers must report crashes involving automated and semi-automated driving systems. With a major update to the National Highway Traffic Safety Administration’s crash reporting order taking effect on June 16, 2025, the gap between Tesla’s FSD track record and that of competitors like Waymo is about to become far more visible to the public and to regulators alike.
New Federal Reporting Rules Raise the Bar
The regulatory backdrop matters because the data that any analyst uses to evaluate FSD safety now flows through a single, more structured federal pipeline. NHTSA’s Third Amended Standing General Order 2021-01, effective June 16, 2025, defines mandatory crash reporting for both Automated Driving Systems and Level 2 Advanced Driver Assistance Systems. Tesla’s FSD falls into the Level 2 ADAS category, meaning every qualifying incident must be disclosed under this framework rather than filtered through company press releases or selective disclosures.
The reporting triggers are specific. Any crash where ADAS was engaged within 30 seconds before impact must be reported, along with required timelines and detailed data elements. That 30-second window is significant because it captures not only moments when the system is actively controlling the vehicle but also situations where a driver disengages the software just before a collision. For Tesla, whose FSD system requires constant driver supervision despite its name, this window could capture a wider range of incidents than the company might prefer to highlight, including edge cases where drivers intervene too late.
The amended order also standardizes how crash data is structured. Reports must include identifiers such as Same Incident ID and Same Vehicle ID, which allow NHTSA and outside researchers to deduplicate crash records and track patterns across manufacturers. That deduplication step is more than a technical housekeeping measure. It means repeated incidents involving the same vehicle, the same geographic location, or the same type of system failure can be flagged systematically instead of being buried in a flood of raw filings that are difficult to reconcile.
For the public, the cumulative effect of these changes is a clearer window into how automated and semi-automated systems behave on real roads. For automakers, the changes narrow the room for ambiguity. A crash that meets the reporting criteria but does not appear in the SGO data will stand out, and a pattern of similar crashes involving the same technology stack will be easier for regulators to spot and investigate.
Waymo’s Data Offers a Sharp Contrast
The analyst’s claim about Tesla’s deteriorating safety metrics gains force when placed alongside the most rigorous public comparison available for a rival system. A recent arXiv preprint analyzed Waymo’s rider-only crash rates across 56.7 million miles of autonomous driving, benchmarking those rates against human driver performance by crash type. The Waymo crashes examined in that study were extracted directly from NHTSA Standing General Order data, making the methodology reproducible and the results directly comparable to what Tesla reports through the same federal channel.
The distinction between the two companies’ systems is not trivial. Waymo operates a fully autonomous, rider-only service with no human driver expected to intervene. Tesla’s FSD is a Level 2 system that explicitly requires a human behind the wheel at all times, with the company emphasizing that drivers must remain attentive and ready to take over. Yet the federal reporting framework treats both under the same umbrella for crash-reporting purposes, which means the public data allows direct side-by-side evaluation on a per-mile basis, at least in principle.
In the Waymo analysis, researchers separated crash types, such as rear-end collisions, intersection conflicts, and single-vehicle incidents, and then compared Waymo’s observed crash rates to human benchmarks derived from national statistics. The study found that for several crash categories, Waymo’s rider-only fleet experienced fewer incidents per million miles than the human baseline. Because the underlying SGO data structure is standardized, a similar breakdown could, in theory, be performed for Tesla’s FSD once researchers isolate the relevant records.
That asymmetry is the core of the analyst’s concern. If Tesla’s FSD crash data is trending in the wrong direction while a fully driverless competitor demonstrates lower rates per mile, the implication is that Tesla’s software updates are not keeping pace with the safety gains its marketing suggests. The federal data pipeline, not any single analyst’s interpretation, is what makes this comparison possible. As more miles accumulate under the amended order, divergence between systems will become more statistically meaningful and more difficult for any company to dismiss as noise.
Why the “Sharply Deteriorating” Claim Deserves Scrutiny
There are real limits to what the available evidence can prove. The analyst’s characterization of Tesla’s FSD safety metrics as “sharply deteriorating” is an interpretation of aggregated federal data, not a conclusion drawn from a controlled study. Tesla has not released its own detailed safety breakdown for recent FSD versions, and no peer-reviewed paper currently isolates Tesla’s post-2024 FSD incident trends from the broader SGO dataset. That gap matters because it leaves room for methodological disagreements over how to count crashes, how to estimate exposure miles, and how to account for software updates that roll out over time.
Without a primary technical analysis comparable to the Waymo arXiv preprint, the deterioration claim rests on secondary readings of the same federal filings rather than on independent, method-transparent research. Analysts may differ on whether to include minor crashes, how to treat incidents where FSD was engaged only briefly, or how to adjust for changes in fleet size and geographic deployment. Small choices in those parameters can produce large swings in apparent trends, especially when the number of reported crashes is still modest relative to total miles driven.
This does not mean the claim is wrong. It means that readers and regulators should treat it as a signal that warrants deeper investigation rather than as a settled conclusion. The NHTSA data is public, and the amended reporting order will make it more granular and more consistent. Any researcher with access to the SGO filings can, in principle, replicate the analyst’s work and verify or challenge the trend. The tools exist. What is missing is a formal, published analysis that isolates Tesla’s FSD crash trajectory with the same rigor that the Waymo study applied to its own system, including clear documentation of assumptions and statistical uncertainty.
What Tighter Reporting Means for Tesla and Drivers
The practical consequences of the updated reporting order extend beyond regulatory compliance. For Tesla owners who rely on FSD for daily driving, the amended rules mean that the federal government will have a clearer, more standardized picture of how the system performs in real-world conditions. If crash rates are indeed rising, that pattern will become harder to obscure as the data grows more structured, more complete, and easier to analyze over time.
For Tesla as a company, the stakes are both regulatory and commercial. NHTSA has the authority to open formal investigations and issue recalls based on patterns identified in SGO data. A visible upward trend in FSD-related incidents could trigger exactly that kind of scrutiny, especially now that the reporting framework is designed to surface repeat failures through vehicle and incident identifiers. The agency’s ability to track whether the same vehicle, the same software version, or the same environmental conditions are involved in multiple crashes adds a layer of accountability that did not exist in earlier versions of the order.
There is also a competitive dimension. Tesla has positioned FSD as the foundation of its future revenue, with plans for a robotaxi service that would compete directly with Waymo and other autonomous fleets. If the federal data consistently shows that Waymo’s fully autonomous system outperforms Tesla’s supervised system on safety metrics, the business case for Tesla’s approach becomes harder to defend to investors, insurers, and municipal regulators who control ride-hail permits. Insurers, in particular, may begin to differentiate premiums based on demonstrable safety records drawn from SGO data, rewarding systems that show lower crash rates over millions of miles.
The Real Test Is About to Begin
The coming months will determine whether the analyst’s warning becomes a footnote or a turning point. With the Third Amended Standing General Order now in force, every qualifying crash involving Tesla’s FSD will enter a public, standardized record alongside incidents from fully autonomous competitors. As that dataset grows, the narrative around FSD safety will shift from dueling anecdotes and selective statistics to a more empirical contest measured in crashes per mile and crash types avoided.
If Tesla’s safety metrics stabilize or improve under the new reporting regime, the company will have a stronger basis to argue that its iterative software updates are delivering real-world benefits. If, instead, the data confirms a pattern of deterioration, regulators will face pressure to act, and Tesla will confront difficult questions about whether its current approach to supervised automation is viable in a marketplace where unsupervised systems can demonstrate superior safety performance. Either way, the combination of tighter federal reporting and independent analysis ensures that the real test of FSD is only beginning, and this time, the results will be much harder to spin.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
